English (United Kingdom)

https://curated-unify.zendy.io/wp-json/zendy-region/v1/featured_content/oa?rat=en

https://curated-unify.zendy.io/wp-json/zendy-region/v1/highlighted_journal/

Global: Zendy Plus annual

Presents the access of premium content as premium feature

Premium Content

Presents the keyphrase highlighting as premium feature

Keyphrase Highlighting

Presents the summarisation as premium feature

Summarisation

Insights

Presents the pdf analysis as premium feature

PDF Analysis

Presents the zaia usage as premium feature

ZAIA

Global: Zendy Plus monthly

Global: Zendy Tools annual

Global: Zendy Tools monthly

Global: Zendy Free Plan

Neural Radiance Fields (NeRF) employ multi-view images for 3D scenerepresentation and have shown remarkable performance. As one of the primarysources of multi-view images, multi-camera systems encounter challenges such asvarying intrinsic parameters and frequent pose changes. Most previousNeRF-based methods often assume a global unique camera and seldom considerscenarios with multiple cameras. Besides, some pose-robust methods still remainsusceptible to suboptimal solutions when poses are poor initialized. In thispaper, we propose MC-NeRF, a method can jointly optimize both intrinsic andextrinsic parameters for bundle-adjusting Neural Radiance Fields. Firstly, weconduct a theoretical analysis to tackle the degenerate case and coupling issuethat arise from the joint optimization between intrinsic and extrinsicparameters. Secondly, based on the proposed solutions, we introduce anefficient calibration image acquisition scheme for multi-camera systems,including the design of calibration object. Lastly, we present a globalend-to-end network with training sequence that enables the regression ofintrinsic and extrinsic parameters, along with the rendering network. Moreover,most existing datasets are designed for unique camera, we create a new datasetthat includes four different styles of multi-camera acquisition systems,allowing readers to generate custom datasets. Experiments confirm theeffectiveness of our method when each image corresponds to different cameraparameters. Specifically, we adopt up to 110 images with 110 differentintrinsic and extrinsic parameters, to achieve 3D scene representation withoutproviding initial poses. The Code and supplementary materials are available athttps://in2-viaun.github.io/MC-NeRF.

MC-NeRF: Multi-Camera Neural Radiance Fields for Multi-Camera Image Acquisition Systems